Skip to content

Commit

Permalink
Simplify locale parser (#5313)
Browse files Browse the repository at this point in the history
This keeps track of slices instead of indices during parsing, and also
simplifies the parser, which removes some checks that were done
redundantly before, which should improve performance.
  • Loading branch information
robertbastian authored Jul 31, 2024
1 parent f5b9401 commit cb02882
Showing 1 changed file with 47 additions and 90 deletions.
137 changes: 47 additions & 90 deletions components/locale_core/src/parser/mod.rs
Original file line number Diff line number Diff line change
Expand Up @@ -17,40 +17,19 @@ pub use locale::{
parse_locale, parse_locale_with_single_variant_single_keyword_unicode_keyword_extension,
};

#[inline]
const fn is_separator(slice: &[u8], idx: usize) -> bool {
#[allow(clippy::indexing_slicing)]
let b = slice[idx];
b == b'-' || b == b'_'
}

const fn get_current_subtag(slice: &[u8], idx: usize) -> (usize, usize) {
debug_assert!(idx < slice.len());

// This function is called only on the idx == 0 or on a separator.
let (start, mut end) = if is_separator(slice, idx) {
// If it's a separator, set the start to idx+1 and advance the idx to the next char.
(idx + 1, idx + 1)
} else {
// If it's idx=0, end is set to 1
debug_assert!(idx == 0);
(0, 1)
};
const fn skip_before_separator(slice: &[u8]) -> &[u8] {
let mut end = 0;

while end < slice.len() && !is_separator(slice, end) {
#[allow(clippy::indexing_slicing)] // very protected, should optimize out
while end < slice.len() && !matches!(slice[end], b'-' | b'_') {
// Advance until we reach end of slice or a separator.
end += 1;
}
// Notice: this slice may be empty (start == end) for cases like `"en-"` or `"en--US"`
(start, end)
}

pub const fn split_out_range(slice: &[u8], start: usize, end: usize) -> &[u8] {
assert!(start <= slice.len());
assert!(end <= slice.len());
assert!(start <= end);
// SAFETY: assertions and align = size = 1.
unsafe { core::slice::from_raw_parts(slice.as_ptr().add(start), end - start) }
// Notice: this slice may be empty for cases like `"en-"` or `"en--US"`
// MSRV 1.71/1.79: Use split_at/split_at_unchecked
// SAFETY: end < slice.len() by while loop
unsafe { core::slice::from_raw_parts(slice.as_ptr(), end) }
}

// `SubtagIterator` is a helper iterator for [`LanguageIdentifier`] and [`Locale`] parsing.
Expand All @@ -67,46 +46,43 @@ pub const fn split_out_range(slice: &[u8], start: usize, end: usize) -> &[u8] {
// All methods return an `Option` of a `Result`.
#[derive(Copy, Clone, Debug)]
pub struct SubtagIterator<'a> {
pub slice: &'a [u8],
done: bool,
// done + subtag is faster than Option<(usize, usize)>
// at the time of writing.
subtag: (usize, usize),
remaining: &'a [u8],
// current is a prefix of remaning
current: Option<&'a [u8]>,
}

impl<'a> SubtagIterator<'a> {
pub const fn new(slice: &'a [u8]) -> Self {
let subtag = if slice.is_empty() || is_separator(slice, 0) {
// This returns (0, 0) which returns Some(b"") for slices like `"-en"` or `"-"`
(0, 0)
} else {
get_current_subtag(slice, 0)
};
pub const fn new(rest: &'a [u8]) -> Self {
Self {
slice,
done: false,
subtag,
remaining: rest,
current: Some(skip_before_separator(rest)),
}
}

pub const fn next_const(mut self) -> (Self, Option<&'a [u8]>) {
if self.done {
let Some(result) = self.current else {
return (self, None);
}
let result = self.subtag;
if result.1 < self.slice.len() {
self.subtag = get_current_subtag(self.slice, result.1);
};

self.current = if result.len() < self.remaining.len() {
// If there is more after `result`, by construction `current` starts with a separator
// MSRV 1.79: Use split_at_unchecked
// SAFETY: `self.remaining` is strictly longer than `result`
self.remaining = unsafe {
core::slice::from_raw_parts(
self.remaining.as_ptr().add(result.len() + 1),
self.remaining.len() - (result.len() + 1),
)
};
Some(skip_before_separator(self.remaining))
} else {
self.done = true;
}
(self, Some(split_out_range(self.slice, result.0, result.1)))
None
};
(self, Some(result))
}

pub const fn peek(&self) -> Option<&'a [u8]> {
if self.done {
return None;
}
Some(split_out_range(self.slice, self.subtag.0, self.subtag.1))
self.current
}
}

Expand Down Expand Up @@ -189,45 +165,26 @@ mod test {
}

#[test]
fn get_current_subtag_test() {
let slice = "-";
let current = get_current_subtag(slice.as_bytes(), 0);
assert_eq!(current, (1, 1));

let slice = "-en";
let current = get_current_subtag(slice.as_bytes(), 0);
assert_eq!(current, (1, 3));
fn skip_before_separator_test() {
let current = skip_before_separator(b"");
assert_eq!(current, b"");

let slice = "-en-";
let current = get_current_subtag(slice.as_bytes(), 3);
assert_eq!(current, (4, 4));
let current = skip_before_separator(b"en");
assert_eq!(current, b"en");

let slice = "en-";
let current = get_current_subtag(slice.as_bytes(), 0);
assert_eq!(current, (0, 2));
let current = skip_before_separator(b"en-");
assert_eq!(current, b"en");

let current = get_current_subtag(slice.as_bytes(), 2);
assert_eq!(current, (3, 3));
let current = skip_before_separator(b"en--US");
assert_eq!(current, b"en");

let slice = "en--US";
let current = get_current_subtag(slice.as_bytes(), 0);
assert_eq!(current, (0, 2));
let current = skip_before_separator(b"-US");
assert_eq!(current, b"");

let current = get_current_subtag(slice.as_bytes(), 2);
assert_eq!(current, (3, 3));
let current = skip_before_separator(b"US");
assert_eq!(current, b"US");

let current = get_current_subtag(slice.as_bytes(), 3);
assert_eq!(current, (4, 6));

let slice = "--";
let current = get_current_subtag(slice.as_bytes(), 0);
assert_eq!(current, (1, 1));

let current = get_current_subtag(slice.as_bytes(), 1);
assert_eq!(current, (2, 2));

let slice = "-";
let current = get_current_subtag(slice.as_bytes(), 0);
assert_eq!(current, (1, 1));
let current = skip_before_separator(b"-");
assert_eq!(current, b"");
}
}

0 comments on commit cb02882

Please sign in to comment.